Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Generate test scenarios, write the test file, and run it for a specific source file using the tailtest R1-R15 rule layer. When the agent needs to (1) cover a file the Stop hook skipped, (2) regenerate tests after refactoring, (3) test a legacy file Codex did not modify this session, or (4) explicitly run tailtest on a named file.
.claude/skills/hashgraph-online-tailtest/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -51% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -62% | 0% |
| case-11 | ✓→✗ | ▼ Worse | -75% | 0% |
| case-14 | ✓→✗ | ▼ Worse | -76% | 0% |
| case-15 | ✓→✗ | ▼ Worse | -37% | 0% |
Generate or update tests for $ARGUMENTS.
Read the source file at $ARGUMENTS. Generate production-like test scenarios covering its public surface -- happy path, key edge cases, and failure modes at the configured depth. Write or update the test file following the tailtest Step 4 rules in AGENTS.md (correct location, correct name, style-matched to existing tests). Run the tests and report only failures; stay silent if all pass.
Treat the file as new-file regardless of its git status -- this skill explicitly requests generation even for legacy files or files the Stop hook would normally skip.
After completing, update .tailtest/session.json: add the file to generated_tests and clear it from pending_files if present.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-03 | fail→fail | 6,028 | 14,475 | +140% | 1 | 1 | 0% | 219 | 351 | +60% | 0 | 0 | — |
case-01 | fail→fail | 9,292 | 16,640 | +79% | 1 | 1 | 0% | 199 | 416 | +109% | 0 | 0 | — |
case-02 | fail→fail | 14,461 | 16,103 | +11% | 1 | 1 | 0% | 166 | 343 | +107% | 0 | 0 | — |
case-04 | fail→fail | 10,518 | 17,793 | +69% | 1 | 1 | 0% | 1,414 | 747 | -47% | 0 | 0 | — |
case-05 | fail→fail | 7,434 | 18,382 | +147% | 1 | 1 | 0% | 1,136 | 609 | -46% | 0 | 0 | — |
case-06 | fail→fail | 11,056 | 2,224 | -80% | 1 | 1 | 0% | 874 | 498 | -43% | 0 | 0 | — |
case-07 | pass→pass | 16,914 | 4,589 | -73% | 1 | 1 | 0% | 2,208 | 897 | -59% | 0 | 0 | — |
case-08 | fail→pass | 11,994 | 9,016 | -25% | 1 | 1 | 0% | 999 | 492 | -51% | 0 | 0 | — |
case-09 | fail→fail | 8,340 | 7,619 | -9% | 1 | 1 | 0% | 532 | 688 | +29% | 0 | 0 | — |
case-10 | fail→fail | 8,316 | 6,945 | -16% | 1 | 1 | 0% | 1,354 | 441 | -67% | 0 | 0 | — |
case-11 | pass→fail | 17,835 | 7,646 | -57% | 1 | 1 | 0% | 2,175 | 549 | -75% | 0 | 0 | — |
case-12 | fail→fail | 3,056 | 15,836 | +418% | 1 | 1 | 0% | 454 | 484 | +7% | 0 | 0 | — |
case-13 | fail→fail | 15,172 | 14,728 | -3% | 1 | 1 | 0% | 1,550 | 2,095 | +35% | 0 | 0 | — |
case-14 | pass→fail | 15,832 | 17,375 | +10% | 1 | 1 | 0% | 2,586 | 627 | -76% | 0 | 0 | — |
case-15 | pass→fail | 12,003 | 18,595 | +55% | 1 | 1 | 0% | 1,207 | 763 | -37% | 0 | 0 | — |
case-16 | fail→fail | 6,472 | 11,231 | +74% | 1 | 1 | 0% | 1,148 | 1,212 | +6% | 0 | 0 | — |
case-17 | pass→pass | 14,625 | 15,900 | +9% | 1 | 1 | 0% | 1,575 | 655 | -58% | 0 | 0 | — |
case-18 | fail→pass | 17,443 | 4,024 | -77% | 1 | 1 | 0% | 1,672 | 640 | -62% | 0 | 0 | — |
case-19 | pass→pass | 8,160 | 8,560 | +5% | 1 | 1 | 0% | 1,515 | 736 | -51% | 0 | 0 | — |
case-20 | pass→fail | 7,318 | 7,018 | -4% | 1 | 1 | 0% | 407 | 506 | +24% | 0 | 0 | — |
case-21 | fail→fail | 18,533 | 16,657 | -10% | 1 | 1 | 0% | 2,298 | 713 | -69% | 0 | 0 | — |
case-22 | pass→fail | 18,477 | 13,831 | -25% | 1 | 1 | 0% | 2,662 | 689 | -74% | 0 | 0 | — |
case-23 | pass→fail | 22,870 | 4,784 | -79% | 1 | 1 | 0% | 3,821 | 414 | -89% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted, and 11 counted toward the lift figure. The other 12 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of -17 percentage points is the difference between those two pass rates over the 11 comparable cases. 8 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.